Inductive Classification of Semantically Annotated Resources through Reduced Coulomb Energy Networks

The tasks of resource classification and retrieval from knowledge bases in the Semantic Web are the basis for a lot of important applications. In order to overcome the limitations of purely deductive approaches to deal with these tasks, inductive (instance-based) methods have been introduced as efficient and noise-tolerant alternatives. In this paper we propose an original method based on a non-parametric learning scheme: the Reduced Coulomb Energy (RCE) Network. The method requires a limited training effort but it turns out to be very effective during the classification phase. Casting retrieval as the problem of assessing the class-membership of individuals w.r.t. the query concepts, we propose an extension of a classification algorithm using RCE networks based on an entropic similarity measure for OWL. Experimentally we show that the performance of the resulting inductive classifier is comparable with the one of a standard reasoner and often more efficient than with other inductive approaches. Moreover, we show that new knowledge (not logically derivable) is induced and the likelihood of the answers may be provided.

Inductive Classification of Semantically Annotated Resources through Reduced Coulomb Energy Networks

The tasks of resource classification and retrieval from knowledge bases in the Semantic Web are the basis for a lot of important applications. In order to overcome the limitations of purely deductive approaches to deal with these tasks, inductive (instance-based) methods have been introduced as efficient and noise-tolerant alternatives. In this paper we propose an original method based on a non-parametric learning scheme: the Reduced Coulomb Energy (RCE) Network. The method requires a limited training effort but it turns out to be very effective during the classification phase. Casting retrieval as the problem of assessing the classmembership of individuals w.r.t. the query concepts, we propose an extension of a classification algorithm using RCE networks based on an entropic similarity measure for OWL. Experimentally we show that the performance of the resulting inductive classifier is comparable with the one of a standard reasoner and often more efficient than with other inductive approaches. Moreover, we show that new knowledge (not logically derivable) is induced and the likelihood of the answers may be provided.

Decision Retrieval and Storage Enabled Through Description Logic

The problem addressed in the paper is how to represent the knowledge associated with design decision models to enable storage, retrieval, and reuse. The paper concerns the representations and reasoning mechanisms needed to construct decision models of relevance to engineered product development. Specifically, AL[E][N] description logic is proposed as a formalism for modeling engineering knowledge and for enabling retrieval and reuse of archived models. Classification hierarchies are constructed using subsumption in DL. Retrieval of archived models is supported using subsumption and query concepts. In our methodology, design decision models are constructed using the base vocabulary and reuse is supported through reasoning and retrieval capabilities. Application of the knowledge representation for the design of a cantilever beam is demonstrated.

The Mathematics of Perception

This chapter summarizes the work on Mathematics of Perception performed by my research team between 2000 and 2005. To support personalization, a search engine must comprehend users’ query concepts (or perceptions), which are subjective and complicated to model. Traditionally, such query-concept comprehension has been performed through a process called “relevance feedback.” Our work formulates relevance feedback as a machine-learning problem when used with a small, biased training dataset. The problem arises because traditional machine learning algorithms cannot effectively learn a target concept when the training dataset is small and biased. My team has pioneered in developing a method of query-concept learning as the learning of a binary classifier to separate what a user wants from what she or he does not want, sorted out in a projected space. We have developed and published several algorithms to reduce data dimensions, to maximize the usefulness of selected training instances, to conduct learning on unbalanced datasets, to accurately account for perceptual similarity, to conduct indexing and learning in a non-metric, high-dimensional space, and to integrate perceptual features with keywords and contextual information. The technology of mathematics of perception encompasses an array of algorithms, and has been licensed by major companies for solving their image annotation, retrieval, and filtering problems.